The Federal Reserve's assets were at $2.26 trillion as of March 3, flat sequentially.
- Securities held outright: $1,971 billion (an increase of $60 billion MoM, resulting from $57 billion increase in MBS and $3 billion in Agency Debt), or a $6 billion decrease sequentially.
- The fed has completed $169.1 billion of $175 billion in the agency MBS program: there is just 3% of Agency dry powder remaining (no new purchases in the week ending March 3). The Fed has completed $1.22 trillion of its $1.25 trillion MBS debt purchase program, or 98%, through March 3 (including the $10 billion announced today). There is now just $35 billion left in Quantitative Easing capacity.
- Net borrowings: unchanged at $103 billion from the prior fortnight.
- Float, liquidity swaps, Maiden Lane and other assets: a $1 billion decline sequentially to $190 billion. The CPFF program was at $7.7 billion. FX liquidity swaps are now at zero. Maiden Lane I and Maiden Lane II increased and were $27.2 and $15.6 billion, while Maiden Lane III came flat at $22.4 billion.
Custody foreign securities holdings increased by $4.4 billion sequentially to $2,969 billion.
The maturity distribution change of Fed assets from the prior week is shown below. Some $6 billion in sub 15 days USTs and Reverse Repos matured, offset by some Agency rolls into this dated category. Other assets maturing up to a year increased by about $8 billion, consisting mostly of Bills rolling closer. The net decline in assets across the entire curve was $6 billion.
The roll off of Quantitative Easing for MBS/Agencies can be seen below. The Fed's ability to artificially keep mortgage rates low is almost over.
And since everyone is concerned about the impact of the end of QE, here is Goldman Sachs to soothe everyone's nerves, by hoping readers believe that as a result of the artifical floow in the MBS market being lifted, the impact will be at most 10 basis point (that's right 10 bps. At least Goldman doesn't see a tightening in rates as a result).
In this Daily Comment we present a statistical analysis suggesting that the impact of the Fed’s asset purchase on mortgage rates has come mostly via the stock of announced MBS purchases rather than the flow of actual purchases. Our estimates suggest a total effect of around 80 basis points (bp), of which 70bp is due to the announced stock of purchases and 10bp to the week-to-week flow of purchases. Taken at face value, our results imply that mortgage rates should only rise by around 10bp when the Fed stops buying MBS over the next few weeks, although the uncertainty remains significant
Many market participants worry that the end of the Federal Reserve’s purchases of mortgage-backed securities (MBS) purchases in March will push up mortgage rates substantially. One popular approach for gauging the likely effect is to estimate the impact of the Fed’s purchase program via the deviation of the 30-year fixed mortgage rate from its normal relationship with the 10-year Treasury yield. As shown in the chart below, mortgage rates are currently about 70bp lower than would be expected based on the current 10-year Treasury yield. (The 30-year mortgage rate is typically compared with the 10-year rather than 30-year Treasury yield because the embedded prepayment option reduces the duration of mortgages compared with Treasuries of equal maturity.) Thus, some analysts have argued that mortgage rates could rise by 70bp once the Fed stops buying MBS.
This approach, however, ignores the distinction between the announced stock of Fed purchases and the flow of actual Fed purchases. If markets are forward-looking and therefore discount the Fed’s eventual demand for assets, rates should depend on the announced stock of purchases. This implies that the entire effect of the Fed’s purchase program should have occurred when it was announced in November 2008 and expanded in mid-March 2009. Only to the extent that markets are imperfect should the mortgage rate respond to actual Fed purchases. The distinction between stocks and flows becomes critical when the Fed stops purchasing MBS but does not change its announced MBS holdings – i.e. what we expect to happen in March.
To asses the outlook for mortgage rates after the Fed stops buying, we construct simple models to estimate the effect of the Fed program on the mortgage rate. It is important to note that it is difficult to be confident in the results of this exercise, as we have few observations and many other factors have affected mortgage rates at the same time. Still, we believe that our results provide a useful benchmark for what might happen to mortgage rates in coming months.
Our model explains the 30-year Freddie Mac primary market fixed mortgage rate (or its spread over 10-year Treasuries) using the stock of announced purchases, the weekly flow of actual MBS purchases, and a number of macroeconomic control variables. The stock of announced MBS purchaseswas initially set at $500bn on November 25, 2008 and then raised to $1.25tr on March 18, 2009. The Fed started purchasing MBS in January 2009; the purchase rate peaked in March 2009 at $33bn per week and has slowed to $11bn per week since then, averaging about $20bn per week during this period. We also include a measure of the corporate Baa yield because we want to allow for changes in overall attitudes to risk that may or may not be related to the Fed’s purchases. Furthermore, we include cyclical indicators such as the ISM manufacturing index, changes in nonfarm payrolls and the unemployment rate. Using a weekly sample from 2007 to the present, we arrive at the following results (see table below):
* First, we consider the effect of the purchase program on the spread between 30-year fixed rate mortgages and 10-year Treasuries, controlling for overall credit spreads as measured by the Baa spread over Treasuries. We estimate that the total effect of the purchase program on the current mortgage/Treasury spread is 70bp; of this, 58bp is due to the announced stock of purchases while 12bp is due to the flow of purchases (see column 1).
* Second, we consider the effect on the level of the mortgage rate, while again controlling for credit conditions by including the Baa yield. We find a total effect of 82bps on the mortgage rate; again most of the effect due to the announced stock (72bp) with the remainder (10bp) due to the flow of purchases (see column 2).
Table: Estimated Effect of the Fed MBS Purchase Program
The magnitude of the total effect we uncover is broadly consistent with a December 2009 speech given by Brian Sack, the head of the New York Fed’s markets group, which implies that Fed purchases have reduced the mortgage rate by about 100bp (see “The Fed’s Expanded Balance Sheet”, 2 December 2009, Remarks at the Money Marketeers of New York University). The magnitudes Sack refers to are broadly consistent with our own previous work. We have argued that Fed asset purchases of $1tr are roughly equivalent to a cut in the federal funds rate of about 100bp or, equivalently, 35bp in financial conditions (see “The Ps and Qs of Unconventional Easing”, Daily Comment, March 18, 2009). Using the weight of the long-term interest rate in the financial conditions index (55%) this effect is equivalent to roughly a 65bp drop in long-term interest (e.g. mortgage) rates for a $1tr purchase program, broadly consistent with the estimates presented above.
While we believe that the Fed’s ultimate objective is to reduce the borrowing rate that consumers actually face – i.e. the level of the primary mortgage rate used in the foregoing analysis – we test the robustness of our findings by considering two alternative measures: (1) a secondary market mortgage rate, and (2) a mortgage rate that is adjusted for the value of the prepayment option (i.e. the fact that borrowers may refinance prior to the maturity of their mortgage). For both of these measures the above conclusion – that the announced stock is the key in pushing down the mortgage rate – continues to hold and the flow effect of the purchases weakens further (not shown in the table).
One key caveat in our analysis is that we only consider the effect of the Fed MBS purchase program on the mortgage rate. We were not able to identify separately the effects of the Fed’s agency debt and Treasury purchase programs on the mortgage rate – essentially because they were announced at the same time and, in the statistical jargon, are hence highly “collinear” with the MBS purchase program. However, to the extent that the agency debt and Treasury purchase programs – which are much smaller in magnitude – had an impact on the mortgage rate, we believe that our approach should implicitly include this effect. In other words, we view our 70-80bp estimate as a reasonable gauge of the total effect of all Fed programs on the mortgage rate. The fact that we find a larger effect on the mortgage rate than on its spread over Treasuries is consistent with a modest impact of Fed purchases not just on mortgage spreads but also on Treasury yields.
In sum, we find that the stock of announced Fed purchases has been more important in pushing down mortgage rates than the flow of actual purchases. Thus, our analysis only points to a modest rise in mortgage rates of around 10bp when the Fed stops buying MBS in a few weeks. Together with the subdued outlook for MBS origination volumes in a weak housing market environment (see “The GSE at a Crossroads”, Fixed Income Monthly, February 2010), this suggests that nominal mortgage rates will remain low. However, an announcement to sell MBS – which we believe will not occur for some time to come – would likely result in a much bigger rise in mortgage rates of up to 80bp. Again, however, we emphasize that it is difficult to have a great deal of confidence in our regression results given the inherent difficulty of estimating the impact of an unprecedented policy in an environment of nearly unparalleled swings in the mortgage and broader asset markets.
Sven Jari Stehn